4.2. Single-Agent Solutions

💡 First Principle: An agent is a generative model given a goal, instructions, and the ability to act — it can call tools and take steps on its own to fulfill a request, not just reply with text. The leap from "chat app" to "agent" is autonomy: the agent decides what to do, where a chat app only generates a response to what you said.

Why care? Agents are new to this exam (they didn't exist on AI-900) and they're explicitly in the syllabus: "create and test a single-agent solution." Expect questions that hinge on whether a scenario needs a plain model call or an agent, and questions about the create-agent-then-talk-to-it workflow.

⚠️ Common Misconception: "An agent is just a chatbot with a fancier name." A chatbot generates text replies. An agent can reason about a goal and autonomously call tools, query data, or take actions to accomplish it. The defining feature is autonomous action toward a goal, not the conversational interface they happen to share.